Ticket #1566 (closed proposed-project: wontfix)

Opened 4 years ago

Last modified 2 years ago

Implement the Immix garbage collector in GHC

Reported by: simonmar Owned by:
Priority: good Keywords:
Cc: mad.one@…, marcot@…, milos.hadzic@… Topic: GHC
Difficulty: 1 person Summer Mentor: not-accepted

Description (last modified by nagnatron) (diff)

Paper:  http://cs.anu.edu.au/techreports/2007/TR-CS-07-04.pdf

The Immix collector by all accounts is a better compromise that either copying or mark/sweep.

It's not a terribly hard job to implement it, but it needs a student that isn't afraid to get their hands dirty in the GC code. I've already done some initial work on it: GHC has a mark-region collector that can reclaim whole blocks, the main difference between this and Immix is that Immix uses a much smaller block size. In GHC we'd have to subdivide blocks and keep track of a liveness bitmap or something - there's a bit of design work to be done anyway.

Comments from Thomas Schilling who looked into this earlier:  http://www.haskell.org/pipermail/cvs-ghc/2009-February/047626.html

Interested Mentors

  • Simon Marlow

Interested Students

  • Austin Seipp <mad.one@…>
  • Marco Túlio Gontijo e Silva <marcot@…>
  • Milos Hadzic <milos.hadzic@…>

Change History

Changed 4 years ago by thoughtpolice

  • cc mad.one@… added
  • description modified (diff)

Changed 3 years ago by marcotmarcot

  • cc marcot@… added
  • description modified (diff)

Changed 3 years ago by nagnatron

  • cc milos.hadzic@… added
  • description modified (diff)

Changed 2 years ago by guest

  • status changed from new to closed
  • resolution set to wontfix

This was a 2010 SoC project. Given that it failed (no Immix GC in GHC, and no prospects of it being added any time soon), it seems unlikely Google will approve another SoC project based on Immix any time soon. Since active tickets should be plausible SoC projects, I'm going to close this.

Note: See TracTickets for help on using tickets.